Overall survival in hematopoietic stem cell transplant patients with hematologic malignancies: Accounting for the unobserved risk factors.

e18553 Background: Allogeneic hematopoietic stem cell transplantation (allo-HSCT) is a potentially curative therapy for hematologic malignancies; it is also associated with severe adverse outcomes including thrombotic microangiopathy (TMA) and mortality. Several studies have reported risk factors associated with these complexities. The objective of our study is to estimate overall survival (OS) for patients with an underlying malignant disease using a joint shared frailty model with TMA. Methods: This is a retrospective analysis of an existing CIBMTR dataset (Epperla et al 2020) that included patients who underwent first allo-HSCT for an underlying malignant disease between the years 2008-2016. All known significant risk factors for OS after allo-HSCT were included as covariates in the joint shared frailty model developed at the disease level. This model allowed for studying survival processes of correlated time-to event outcomes death and TMA. Frailty, a latent variable accounting for heterogeneity between patients, captures the shared risk that affects both the outcomes (mortality and TMA) in a similar way within each patient. Patients were classified into three risk groups based on the distribution of predicted frailty scores (FS) —low (FS < Q1), moderate (Q1 < FS < Q3) and high (FS > Q3) risk. Analyses were performed using frailtypack, an R package and SAS 9.4. Results: A total of 20095 allo-HSCT patients with an underlying malignant disease, with 10142 (50.5%) deaths and 561 (2.8%) TMA cases were included in the analysis. Frailty effect, the combined effect of unobserved covariates, was significant (p<0.0001) along with other previously reported significant risk factors for mortality and TMA. One-year OS and median survival (MS in months) from HSCT of low (N=5077), moderate (N=10011), and high-risk (5007) patients was 99.6% (not estimable), 73.1% (26.6) and 8.0% (3.3), respectively, as compared to observed 63.7% (31.3). One-year OS and MS (in months) from TMA diagnosis of low (N=141), moderate (280), and high-risk (140) patients was 77.0% (88.5), 38.8% (6.2) and 5.2% (1.2), respectively, as compared to observed 41.0% (5.9). Overall survival from HSCT and from TMA were significantly (p<0.0001) inferior (HR 13.5, 95% CI12.9-14.1 and HR 4.7, 95% CI 3.8-6.0, respectively) for the high-risk patients as compared to the low and moderate risk patients. Conclusions: Our study highlights the importance of accounting for the effect of unobserved risk factors in patients, while assessing all observable covariates. Predicting frailty scores is crucial for identifying patients who may need more intensive monitoring and intervention to avoid adverse effects following allo-HSCT. Future avenues of research include possibly developing an interactive shiny app to enable clinicians to estimate individualized survival, decision-making and risk mitigation.